LLaMAntino-3-ANITA-8B-Inst-DPO-ITA — benchmark results
University of Bari's (swap-uniba) Italian-English fine-tune of Llama 3 8B Instruct, QLoRA SFT plus DPO alignment from the ANITA project. Provider: Meta. Released 2024-04-29. Access: Open.
Unified ELO 1442 ± 10, rank #1036 of 1776 rated models, from 25 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| EVALITA - faq | 56.55 | CPS | 100 |
| EVALITA - relation-extraction | 42.14 | CPS | 87.2 |
| Open Italian LLM - MMLU-Pro (IT) | 30.11 | Accuracy (%) | 85.7 |
| Open LLM Leaderboard - MuSR | 13.24 | Score | 72.9 |
| EVALITA - summarization-fanpage | 29.1 | CPS | 72.3 |
| Pinocchio Italian - Logica | 43.97 | Accuracy (%) | 70.5 |
| EVALITA - hate-speech-detection | 70.99 | CPS | 70.2 |
| Pinocchio Italian - Cultura | 60.4 | Accuracy (%) | 65.9 |
| Pinocchio Italian - Diritto | 54.65 | Accuracy (%) | 65.9 |
| Pinocchio Italian - Generale | 57.71 | Accuracy (%) | 65.9 |
| Pinocchio Italian Leaderboard | 55.38 | Average Accuracy (%) | 65.9 |
| Open LLM Leaderboard - MMLU-Pro | 30.26 | Score | 61.2 |
Interactive version: theaggregate.ai/model?slug=llamantino-3-anita-8b-inst-dpo-ita · How the rankings work · Data refreshed daily, snapshot 2026-07-22.